Disentangling the dust and gas contributions of the JWST/MIRI spectrum of Sz 28

Context. Recent spectra of protoplanetary disks around very low-mass stars (VLMS), captured by the Mid-InfraRed Instrument (MIRI) on board the James Webb Space Telescope (JWST), reveal a rich carbon chemistry. Current interpretations of these spectra are based on 0D slab models and provide valuable estimates for molecular emission temperatures and column densities in the innermost disk (radius ≲ 1 au). However, the established fitting procedures and simplified models are challenged by the many overlapping gas features. Aims. We aim to simultaneously determine the molecular and the dust composition of the disk around the VLMS Sz 28 in a Bayesian way. Methods. We modelled the JWST/MIRI spectrum of Sz 28 up to 17 μm using the Dust Continuum Kit with Line emission from Gas (DuCKLinG). Systematically excluding different molecules from the Bayesian analysis allowed for an evidence determination of all investigated molecules and isotopologues. We continued by examining the emission conditions and locations of all molecules, analysing the differences to previous 0D slab fitting, and analysing the dust composition. Results. We find very strong Bayesian evidence for the presence of C 2 H 2 , HCN, C 6 H 6 , CO 2 , HC 3 N, C 2 H 6 , C 3 H 4 , C 4 H 2 , and CH 4 in the JWST/MIRI spectrum of Sz 28. Additionally, we identify CH 3 and find tentative indications for NH 3 . There is no evidence for water in the spectrum. However, we show that column densities of up to 2 × 10 17 cm −2 could be hidden in the observational noise if assuming similar emission conditions of water as the detected hydrocarbons. Contrary to previous 0D slab results, a C 4 H 2 quasi-continuum is robustly identified. We confirm previous conclusions that the dust in Sz 28 is highly evolved, with large grains (5 μm) and a high crystallinity fraction being retrieved. We expect some of the stated differences to previous 0D slab fitting results to arise from an updated data reduction of the spectrum, but also due to the different modelling process. The latter reason underpins the need for more advanced models and fitting procedures.